Prompts

Actions

Prompts

• Manage
• Suggestions

Actions

Prompts

• the prompts and topics currently tracked
• proposed changes waiting for your decision

Overview#

This is the control room for everything Seerly measures. Your topics, and the prompts sitting inside them, decide which questions get asked on every run, which means every visibility, sentiment and citation number in the product traces back to decisions made on this screen.

Your first set was created for you during setup. This is where you keep it accurate as the business changes around it, and it is worth revisiting more often than most people do.

Manage#

All prompts lists everything currently tracked, and Discovery topics are the ones derived from your keyword map when the account was set up.

Add Topic creates a new tracked topic and Add Prompt adds a question to one. Where a question belongs somewhere you already track, Add Prompt to Existing Topic or Add to existing topics files it there instead of spawning a near-duplicate topic alongside the original, which is worth doing deliberately: two topics covering the same ground split your data across two rows that should have been one, and neither row then has enough behind it to be stable. Add another prompt keeps you going without leaving the screen.

All prompts

Prompt	Avg Pos	Citation Share
What's the best project management tool for a small agency?	2.1	18%
Is paid project management software worth it for five people?	4.6	7%

Chained prompts#

Chain (add follow-up steps) and Add step let you build a prompt that is a short conversation rather than a single question.

This exists because real buyers very rarely ask one question in isolation and then act on the answer. They circle a subject, give some context about themselves, and then ask the thing they actually wanted to know. A cold question and the same question asked once context has been established can produce genuinely different answers, and chained prompts are how you measure the second case. If you want to try a chain out before committing it to your tracked set, Primed Prompts is the place to experiment without changing anything.

Change History#

Change History is an audit trail of what has been changed on a prompt, described on screen as exactly that. No change history found means nobody has edited it since it was created.

It is more useful than it sounds. When a topic's numbers shift and nobody can account for it, the explanation is very often that somebody rewrote a prompt three weeks ago and the question being asked is no longer quite the question that produced the earlier trend. Checking here first will save you investigating a market that never moved.

You may also run into No prompts found or No topics found where nothing matches.

Suggestions#

ActionsPromptsSuggestions

These are proposed changes to your prompt set, produced by the Brand Monitoring Scout agent, which looks for questions you ought to be tracking and are not.

Each proposal is either accepted or Dismissed, and Generated Prompts are the ones it has drafted for you. No proposed changes in this run means it found nothing worth raising, which is a healthy outcome rather than a failure of the agent.

If the Scout is not running you will see a prompt to Deploy your Brand Monitoring Scout, with Go to Agents taking you to Agents to set it up.

What a suggestion run can say

No changes proposed

a healthy outcome, not a failure

Applied

you accepted the run's changes

Partially applied

you took some and dismissed the rest

Outdated

a newer run has superseded this one

Dismissed

declined, and the reason teaches the Scout

Dismiss things with intent while you are here. A dismissal teaches the Scout what you do not want, so a considered dismissal is worth appreciably more to you later than letting the list accumulate unread.

Best practices#

• Topics that describe what you actually sell, reviewed every so often rather than set once and forgotten.
• Enough prompts per topic to be stable. With too few, results swing between runs on ordinary variation rather than on anything meaningful, and you end up reacting to noise.
• A spread of question kinds rather than only "best X for Y". Comparison questions and experience questions expose completely different weaknesses, and you can be strong in one kind while absent from another.
• Prompts phrased the way buyers actually speak. Read them aloud. Anything that sounds like a keyword rather than a question will produce answers that bear little resemblance to a real buying conversation.
• Suggestions reviewed rather than left to pile up.

Next steps#

1. Review your topics before trusting any metric anywhere else. If they are wrong, everything downstream is a careful description of somebody else's business.
2. Add prompts for questions you know buyers ask and that are missing. You know things about your market that no extraction process is going to infer from your website.
3. Prefer adding to an existing topic over creating a near-duplicate.
4. Try a chain for anything where context plausibly changes the answer.
5. Check Change History when a number moves unexpectedly, before you go looking for a market explanation.
6. Work through Suggestions regularly, dismissing with intent so the Scout gets better at proposing.

Limitations#

Prompt volume is capped by plan, both per topic and in total. See seerly.app/pricing.

Adding a prompt starts measuring from the next run onward. It does not backfill history, so a new prompt begins its own trend rather than filling in a past it was never part of.

A question that no prompt represents is not measured at all, and an unmeasured gap looks exactly like no gap from every other screen in the product. Coverage of your prompt set matters as much as any score it produces.